ASA Connect

 View Only

  • 1.  Effect Size for Intraclass Correlation Coefficient

    Posted 04-09-2018 08:40
    Hi ASA Community,

    Any  recommendations for articles about effect size guidelines for intra-class correlation coefficients?
    According to Cohen (1992), a Pearson correlation coefficient of .1 has a small effect size, .3 is a medium effect size, and .5 is a large effect size.  I would appreciate any articles or other references about effect sizes for intraclass correlation coefficients.

    Reference mentioned above:
    Cohen, J. (1992). A power primer. Psychological Bulletin, 112(1), 155-159.

    --
    Regards,
    Brandy R. Sinco
    Statistician and Programmer/Analyst, UM School of Social Work
    Current Projects: 
    Mon, Wed, Fri CHW Integration/REACH; 
    Tues, Thurs  RISE/WCBT


  • 2.  RE: Effect Size for Intraclass Correlation Coefficient

    Posted 04-10-2018 16:19
    Most of the publications on power/sample size/effect size analysis use Jacob Cohen's classic: Statistical Power Analysis for the Behavioral Sciences, Revised 2, 2013 by Routledge (original: 2nd Edition, Lawrence Erlbaum Associates, 1988), Can be read online at www.amazon.com/statistical-power-analysis-behavioral-sciences/dp.

    There is an article by Chao-Ying Joanne Peng, Haiying Long, and Serdar Abaci, Power Analysis Software for Educational Researchers, from Indiana University that summarizes recent articles and software collections with references.  Can get a copy from Peng@indiana.edu.  It is an excellent reading.

    As for software collections, most of them are based on Cohen's algorithms:

    1. GPOWER 2.0 (or its Macintosh version) developed by Edgar Erdfelder, Franz Faul, & Axel Buchner, GPOWER. A General Power Analysis Program, Behav. Res. Methods, Instruments, & Computers, 28(1):1-11, 1996.  Free. Can be downloaded from the public FTP server at the University of Trier, Germany.  It is based on Cohen's book.  

    2. PASS 15 by NCSS Software, Commercial.  www.ncss.com.  The father and son Ph.D. applied statisticians developed and maintain it which is based on Cohen's book.

    Finally, if one is also interested in exact computation of power/sample size/effect size computation, there is CyTel Corp.'s East 6.3.  It will do asymptotic, Monte-Carlo, and exact computations. It is commercial.  Contact: www.cytel.com/software/east.

    Many university statistics/biostatistics centers (e.g. University of Michigan, University of Wisconsin, etc.) offer free software developed for these purposes.

    Hope you find what you are looking for in the above.

    Ajit K. Thakur, Ph.D.
    Retired Statistician






    Original Message


  • 3.  RE: Effect Size for Intraclass Correlation Coefficient

    Posted 04-10-2018 20:27
    Hello Brandy, firstly of course Pearson's correlation measures association not agreement.  The ICC and Kappa measure agreement.  The ICC via ANOVA for continuous measures can measure within variance agreement for laboratory serum measures for example.   Kappa via contingency tables can for example measure rater agreement on some discrete scale.

    See Bartko On the Methods and Theory of Reliability.  The Journal of Nervous and Mental Disease,Vol 163 No 5 1976 pages 307-317.

    There are algebraic mappings between the ICC and Kappa.

    For ICC effect sizes see for example Cicchetti, Domenic 1994 Psychological Assessment 6 (40 284-290.

    For Kappa effect sizes see Landis and Koch 1977 The measurement of observer agreement for categorical data Biometrics 33, 159-174.

    See Also Fleiss's book on Rates and Proportions, Wiley.

    The above just skims the surface but you may find something for your use here.




    All the best.   John Bartko


    ------------------------------
    John Bartko
    Consulting Biostatistician
    ------------------------------

    Original Message


Related Content